In one embodiment, a two-dimensional to stereoscopic conversion method, comprising: estimating a local motion region in a first image relative to one or more second images, the first and the one or more second images comprising two-dimensional images; generating a color model based on the local motion region; calculating a similarity value for each of at least one image pixel selected from the first image based on the color model; and assigning a depth value for each of the at least one image pixel selected from the first image based on the calculated similarity value to generate a stereoscopic image, the method performed by one or more processors.